Return Intentions of Irregular Migrants and Skills:the Perverse Effect of the Brain Waste∗

Nicola D. Coniglio† Giuseppe De Arcangelis‡ Laura Serlenga§

Abstract

In this paper we show that highly skilled illegal migrants may be morelikely to return home than migrants with low or no skills when illegalitycauses “brain waste”, i.e. lower ability of exploiting individual abilitiesboth in the labor and the financial markets. This result is in contrast withcommon wisdom on return migration, according to which low-skill individ-uals are more likely to go back to the home country rather than high-skillmigrants. The theoretical model is tested with a data set on a sample of920 illegal migrants crossing Italian borders in 2003. The estimation resultsconfirm that highly skilled illegal migrants are more willing to return home.Moreover, a higher probability of being granted legal status (as it happensfor asylum seekers rather than clandestines) lessens the intention to return.

Keywords: Illegal migration, brain waste, labor skills.JEL Classification Codes: F22, C25.

∗We wish to thank Maria Concetta Chiuri and Giovanni Ferri for useful discussions on thepreliminary versions of this paper. Piero Cipollone and Christian Dustmann also offered usefulinsights at the AIEL 2005 Meeting (Rome 22-23 September 2005) where the paper has beenpresented. The usual disclaimer applies.

†Norwegian School of Economics and Business Administration (NHH) and University of Bari.‡University of Bari and Univesity of Rome “La Sapienza”.§CREST and University of Bari.

1

1 Introduction

The debate on illegal migration in the developed world is capturing a great dealof public attention. The mounting dimension of the phenomenon is a direct con-sequence of the tightening of immigration laws in most OECD countries. In fact,instead of decreasing the size of immigration flows, this policy is having the effectof increasingly shifting the balance from legal to illegal migration.1 In terms ofeconomic and social impact for both receiving and sending countries this patternis far from being neutral. Given the different set of incentives and constraintsfaced by legal and illegal migrants we might easily expect significant differencesin their migratory behavior. Nevertheless, while there are numerous contributionsin the literature on legal migration, the phenomenon of illegal migration has beenscarcely analyzed, mainly because of the severe lack of data.

In this paper, we aim to shed some light on return migration, and in particularon return decisions of illegal migrants. Generally, return migration is importantfor both the country of origin and the host country. Since return migrants mostlycarry capital, knowledge and entrepreneurship to developing countries, countries oforigin are interested in understanding both the determinants of the return choiceand the individual characteristics of those who decide to return. On the otherhand, in order to establish well-designed immigration policies, the analysis of theindividual behavior of migrants (i.e. getting information on plans and future ex-pectations) is also essential for destination countries. This is valid for both legaland illegal migrants; in particular, the latter ones must consider both the possi-bility of being apprehended, but also the probability of being granted legal status(e.g. as asylum seeker or for a general amnesty).

Our analysis focuses on illegal migrants for whom illegality is originally designedin economic terms as a brain waste effect, i.e. a tax that impinges the positiveoutcome of skills on both individual income and savings. In particular, we analyzetwo main aspects of the choice of return for illegal entrants: the effect of individualskills and the effect of possibly being granted legal status.

Regarding the relationship between skills and return intentions, most literaturehas focused on legal migrants. Many studies have emphasized that migrants are notrandomly selected but generally represent the upper tail of the skills distributionof the population in the countries of origin (see Borjas et al. 1992; Chiswick 2005).Since migration is a particularly costly investment, only the most capable, entre-preneurial and risk prone individuals usually undertake such an investment. The

1According to recent estimates of the INS the total unauthorized immigrant population resid-ing in the United States in January 2000 was about 7.0 million; from 1990 to 1999 from 350.000to 500.000 illegal migrants where crossing the US borders annually. Estimates of illegal migrationflows to Europe (EU-15) in 2001 are up to 650.000 according to a recent study by Jandl (2003),(100.000 of them in Italy).

2

existing empirical research almost unanimously concludes that return migration ismore likely for individuals with low skills and reinforces the positive self-selectionof the migrants (Borjas et al. 1996; Dustmann 1993, 2003a, 2003b; Reagan andOlsen, 2000).

However, given the different constraints that legal and illegal entrants face,what is found to apply to the legal migrants may not apply to illegal ones. Asgenerally acknowledged, although one of the most common motives for migrationis the necessity to accumulate assets (which will be subsequently employed inproductive activities) an illegal entrant is generally less capable to fully exploit hisor her skills and human capital. Moreover, the illegal status hinders the migrant’saccess to many markets and institutions in the host country (including financialmarkets), which are instead fully available to legal migrants. Being illegal maymake individual skills even less effective than in the home country, as the illegalmigrant has to resort uniquely to the shadow economy. As a consequence, thebrain waste effect, typically related to the illegal status, is particularly strong forthose who are the most skilled and educated among the illegal entrants. Giventhis, it would be natural to expect that the opportunity cost of returning to thecountry of origin be substantially lower for the skilled individuals than for theunskilled ones.

With respect to the probability of being granted legal status and its conse-quences on the return choice, this is strictly applicable to illegal migrants. Indeed,it is important to distinguish between two broad classes of illegal migrants, asylumseekers and clandestine immigrants, which differ in two important respects: theirdesire/ability to live in the open vs. staying hidden; their wish/faculty to return tothe home country vs. residing permanently in the country of immigration. Asylumseekers are motivated to notify their presence to the authorities of the receivingcountry, whereas clandestine immigrants shy away from official contacts and tendto live working quietly. As to return migration, this is an option open to clandes-tine immigrants but instead generally unavailable to asylum seekers, at least untilmajor events change the situation in the country of origin. While both asylumseekers and clandestine immigrants face the real risk of repatriation, for the lattergroup this risk is more pronounced.2 Different probabilities of being expelled to-gether with different incentives for being “visible” in the country of destination willhave a likely effect on labor market performance of illegal migrants, for instancetheir ability to gain good employment opportunities and length of unemploymentspells. The effect on the return choice of different probabilities of being grantedlegal status is studied in this paper by comparing the distinctive attitudes of (the

2On one hand, clandestine immigrants will be generally repatriated upon apprehension, anevent that might materialize with some positive probability. On the other hand, the outcome ofthe generally long and complex procedure following their request might be unfavorable to asylumseekers, in which case they would also be repatriated.

3

high-probability) asylum seekers vs. (the low-probability) clandestine immigrants.The contribution of this paper is twofold. First, we provide a theoretical model

that links skills, probability of getting legal residence and magnitude of the brainwaste effect to the return decision of illegal migrants. Second, thanks to theavailability of an unique data set, we empirically test the main implications of ourtheoretical specification.

In the theoretical model, the illegal migrant is already in the host country andhas to decide whether to stay or to return in the second period. Average wages andrates of return are higher in the host country, but the illegal migrant must bearthe cost of illegality, i.e. the brain waste effect that affects relatively stronger high-skill individuals rather than low-skill ones. Average wages and financial returns arelower in the home country, but no brain waste occurs. This trade off characterizesthe return choice of the illegal migrant, together with the probability of beinggranted legal status in the second period and avoid the brain waste effect.

We mainly find that: (i) the probability of returning in the home country is anincreasing function of the individual level of skills; and (ii) an increasing probabilityof being granted a legal status has a negative effect on the return propensity.

These implications are tested using a data set on a sample of 920 illegal mi-grants who crossed the Italian borders in 2003. One of the most important featuresof these data is that they contain information on the migrants’ expectations “atthe gate” concerning their intentions to return, together with many other char-acteristics (e.g. intentions to remit, expectations on future income, employment,legal status, characteristics of the village of origin etc.). Indeed, using this dataset we are able to quantify the effects of skills and education and other relevantvariables on the return choice.

Empirical results confirm the main findings of our theoretical model and, inparticular, they highlight the important role of individual skills in increasing thechoice of return for illegal migrants. Also, the possibility of being granted a legalstatus has a significant role on the probability of return.

To the best of our knowledge, this is the first contribution towards increasingour knowledge on the relationship between skill characteristics and return attitudesfor illegal migrants, whose numbers far outpace those of legal migrants.

The paper is organized as follows. Section 2 presents the theoretical modelof return plans of irregular migrants with heterogeneous level of skills. Section 3describes the main characteristics of the data set employed to test the theoreticalmodel. Section 4 reports and discusses the results of the empirical analysis. Lastly,Section 5 concludes with some general remarks and suggestions for further research.

4

2 The model

Consider a population of illegal migrants with a heterogeneous level of skills fromthe same source country A who have migrated to the host country B. Migrants’skills are continuously distributed over an interval [a, a] where a and a representrespectively the individuals with the lowest and the highest skill level.

Individuals operate in a two-period world and are endowed with a unit of laborwhich is inelastically supplied in each of the two periods.3

The migrants’ intertemporal utility function is defined over first- and second-period consumption and takes the following simple form:

U (c1, c2) = u (c1) + δu (c2) = ln (c1) + δ ln (c2)

where δ is the discount factor.In the first period individuals live and work in the host country B. Consump-

tion of migrant j is:

cj1 = wj

1 − sj

where wj1 is the first-period wage when working illegally in country B and sj are

savings.Given their status of illegal migrants in the host country B the rewards to

human capital cannot be fully exploited: income earned in country B is increasingin the skill level but we assume that the skill premium is compressed because ofillegality. More precisely, first period wages are given by the following equation:

wj1 = ajτwB

where wB is the exogenously given “normal” wage for a unit of labor in the hostcountry.

Individual wages positively depend on individual skills but the status of illegalmigrant makes those skills less effective. The parameter τ ∈ (0, 1] captures themagnitude of the brain waste effect associated with the status of illegal migrant. Asτ → 0 illegal migration tends to be less and less rewarding for all illegal migrantsand has a squeezing effect on the level of human capital, i.e. being uneducated andunskilled rather than having a PhD in engineering does not change the returns frommigration.4 On the contrary, when τ = 1 there is no brain waste and migrants’

3We assume that the individual possesses no capital at the beginning of the first period. Inreality, it is often the case that migrants from less developed countries have a negative amountof wealth since they have borrowed from friends and relatives in order to pay for migration costs.

4Even if τ = 0 is implausible since the brightest and more skilled migrants are more likely toobtain the best opportunities, skills and formal qualification are of little use if you are an illegalmigrant. Very often migrants employed illegally in highly unskilled and manual jobs – such as

5

human capital is fully rewarded according to the skill content aj.5In other words,when τ = 1 we assume that migration is legalized.

The parameter τ might be interpreted as the effect of the institutional frame-work within which illegal migration takes place on the individual’s ability to usethe stock of human capital accumulated at home. The degree to which it is possi-ble for the migrant to exploit his or her skills might depend, for instance, on theattitude of the immigration authorities in the host country. When some particularskills are required due to an excess demand in the host country labor market, im-migration authorities tend to be more tolerant toward illegal migrants possessingthose skills (in this case τ may be close to 1).

In the second period migrants face two options. They can return to the homecountry A, where the exogenously given “normal” wage is wA (< wB). In this casethey fully use their skills and earn ajwA. Alternatively, they continue to residein the host country B where they face a positive probability of becoming legalmigrants and therefore fully exploit their human capital.

The brain waste affects also the ability of illegal migrants to fully exploit finan-cial markets in the host country and therefore the return on savings, which differsdepending on the migrant’s choice for the second period.

Often the sole motive for migration is the necessity to accumulate assets thatwill be subsequently employed in productive activities at home. Here we assumethat if the migrant decides to go back to homeland A in period 2, then period-1savings will be directly used, together with individual skills, in an entrepreneurialproject with gross return ajRA in the home country A — where RA is the exoge-nously given “normal” gross return on savings in the home country. We allow forreturns from the entrepreneurial project to differ between migrants. The higherthe level of skills of the migrant, the higher the likelihood that she will locatethe best investment opportunities and, in turn, the more rewarding will be theallocation of her capital.

Similarly, savings are located in the host country B in case the migrant decides

agricultural workers in developed countries – are highly skilled and educated individuals.5Since all individuals found it profitable to migrate at the beginning of the first period and

given that we abstract from differences in preferences for the location of consumption (associatedfor instance with relatively high preferences for home consumption) for any aj ∈ [a, a] thefollowing inequality is satisfied:

τajwB ≥ ajwA =⇒ τwB − wA ≥ 0

where wA(< wB

)is the exogenously given “normal” wage for a unit of labor in the home country.

In other words wage differentials more that compensate for the “brain waste” effect. Moreover,since we assume that illegal migrants have already chosen to live and work in the host countryB in period 1, the condition above imposes either a lower bound to the percentage wage gapw ≡ wB

wA (i.e. w > 1τ ) or, given wA and wB , a lower bound to τ (i.e. τ > wA

wB ).

6

to stay in B during period 2. The exogenously given “normal” return on savingsin B is RB. Then, in case of a period-2 stay in country B, savings generate areturn τajRB, which is higher for individuals with higher skills, but is affected bythe brain waste.

Hence, the return from savings will vary according to the migrant’s locationchoice for the second period:

ej =

{ej

R = ajRAsj

ejNR = τajRBsj

if he or she returns to country Aif he or she stays in country B

In other words, illegal migrants face constraints which negatively affect notonly their ability to fully exploit their labor potential but also their ability tolocate and exploit investment opportunities. For instance, although fully aware ofthe different financial opportunities offered in the host country, the illegal migrantdoes not have access to them since she does not have a legal permit and must recurto alternative, less rewarding and sometimes illegal, forms of financial investment.Instead, when planning to go back to the homeland, migrants immediately sendhome their savings, where they start their entrepreneurial project even beforereturning.

Therefore, consumption in second period also differs depending on the mi-grant’s second-period choice. In case of return migration, consumption is givenby:

cj2,R = wj

2,R + ejR = ajwA + ajRAsj = aj

(wA + RAsj

)where in the home country return migrants are fully able to exploit their humancapital as related to both their endowment of labor and the capital saved in thehost country.

If migrants decide to stay in the host country they face a positive probability ofgetting legal residence. For instance, this might happen in the case of an amnestygranted to all illegal migrants who have being residing and working for a certainperiod in the host country or in the case of acceptance of an asylum application.The main consequence of being granted legal status in terms of our model is theability to fully make use of individual skills, i.e. the brain waste effect disappearsin the second period when the migrant obtains the legal status.

Consumption in this case can be expressed as the expected income in period

2 (wj2,NR ≡ wj,B) plus the accumulated savings, invested in the host country B

(ejNR):

cj2,NR = wj

2,NR + ejNR = wj,B + ej

NR (1)

Given γ as the probability of getting legal residence in period 2, then the

expected wage for migrant j in country B in period 2 (wj,B) will be: (i) τajwB,

7

i.e. the illegal immigrant’s wage (the same as in period 1) in case of not gettinglegal status, with probability (1− γ); (ii) ajwB, i.e. the legal immigrant’s wage incase of getting legal residence, with probability γ.

Hence, the expected wage for period 2 in case of no-return is:

wj,B = (1− γ)τajwB + γajwB = ajhwB = ajwB

where h ≡ [(1 − γ)τ + γ] and wB is the expected “normal” period-2 wage in thehost country B.

When substituting both expected income for period 2 and the return on savinginto the expression (1) for consumption, it yields:

cj2,NR = ajhwB + τajRBsj = ajwB + τajRBsj

Finally, the lifetime utilities functions of migrants depend on their decisionwhether or not to return. In the case of return:

U jR (c1, c2) = ln

[τajwB − sj

]+ δ ln

[aj

(wA + RAsj

)](2)

Whereas in the case of no return:

U jNR (c1, c2) = ln

[τajwB − sj

]+ δ ln

[aj(wB + τRBsj)

](3)

In the following sections we compute the optimal level of savings in both casesand focus on the relationship between the illegal migrant’s skills level and herrational decision whether or not to return to the home country.

2.1 Optimal Savings, Return Decisions and Skills

The optimal level of savings s∗j for an individual with skills j is conditional on herlocation decision for the second period.

In the case of return migration the level of savings which maximizes the indi-vidual’s intertemporal utility function (2), is given by:

sj,∗R =

1

RA(1 + δ)

[δRAwj

1 − wA]

=1

RA(1 + δ)

[δτajRAwB − wA

](4)

If the illegal migrant decides to stay in the host country, then the optimal first-period savings will be determined by the maximization of the utility function (3).Hence, the optimal savings in case of no return is the following:

sj,∗NR =

1

τRB(1 + δ)

[δτRBwj

1 − wB]

=wB

τRB(1 + δ)

[δτ 2ajRB − h

](5)

8

since wB ≡ hwB and h ≡ [(1− γ)τ + γ].It is easy to show that savings in case of return are higher than saving in case

of no-return when the percentage wage gap between the host country B and theorigin country A — that is w ≡ wB

wA — is higher than the percentage rate-of-return

gap — that is R ≡ RB

RA — i.e. when w > R.6

Several authors have emphasized that a positive probability of return inducesmigrants to save and remit more (see Galor and Stark, 1990; Stark, 1992; Mesnard,2004). This result is in accordance with the life-cycle theory of consumption sinceindividuals who plan to re-emigrate in a relatively poor country will save more inorder to smooth their consumption path over the life-cycle.7

By substituting the optimal level of savings (4) and (5) in the respective utilityfunctions (2) and (3), we obtain the optimal levels of utility in case of return (U j,∗

R ):

U j,∗R (δ, τ, aj, wA, wB, RA) = (1 + δ) ln

[1

1 + δ

(RAτajwB + wA

)]−

− ln(RA) + δ ln(δaj) (6)

and in case of no-return (U j,∗NR):

U j,∗NR(δ, τ, aj, wB, RB) = (1 + δ) ln

[wB

1 + δ

(RBτ 2aj + h

)]−

− ln(τRB) + δ ln(δaj) (7)

Let us define the net utility derived from returning U j,∗ for an illegal migrantwith j level of skills as the difference between the two optimal levels of utility.Hence:

U j,∗(δ, τ, aj, wA, wB, RA, RB) ≡ U j,∗R − U j,∗

NR ≡

≡ (1 + δ) ln

[RAτajwB + wA

τRBτajwB + hwB

]− ln

RA

τRB(8)

which can be rewritten as:

6More precisely, sj,∗R > sj,∗

NR when:

w

R>

τ

[(1− γ)τ + γ]

Note that the fraction τ[(1−γ)τ+γ] is always lower than 1 since τ ∈ (0, 1].

7Higher incentives to save could also be motivated by a higher marginal utility of consumptionin the home country, for instance due to higher purchasing power in the home country or strongpreferences for home varieties or by the necessity to overcome higher uncertainty (see Dustmann1997).

9

U j,∗ ≡ (1 + δ) ln

[RAwj

1 + wA

τRBwj1 + wB

]− ln

RA

τRB(9)

The discrete choice whether or not to return depends on the sign of the unob-servable net utility U j,∗.

Propositions 2 and 3 in the Appendix A show the sufficient conditions on wagesand rate of returns such that the sign of U j,∗ is always negative or always positive,i.e. such that migrants respectively decide never to return or always to return.

However, such sufficient conditions are not easy to satisfy. For instance, in thespecial case of equal “normal” rates of returns — RA = RB — the Propositions 2and 3 never hold when the “normal” wage in the host country is greater than the“normal” wage in the home country, i.e. wB > wA, which is the most commoncase also in reality.

Instead, when migrants are able to circumvent the effect of the brain wasteonly in the financial markets and rates of return are equalized net of the brainwaste — i.e. RA = τRB — then the decision where to work depends exclusivelyon the total flow of income in the two locations. Since the migrant starts in thehost country under both cases, the decision regards only income from period 2.

The migrant decides (not) to return if and only if: wA > wB (wA < wB), with norole played by the individual skills.

Notwithstanding these special instances, in the most general case the twopropositions show that commonly the sign of U j,∗ is not uniquely defined. Amongall the parameters that denote the sign of the net utility, we pay particular atten-tion to the skill content, represented by aj.

In particular, the derivative of the net utility U j,∗ with respect to aj is thefollowing:

∂U j,∗

∂aj=

(1 + δ)τwB

WRWNR

(hRAwB − τRBwA

)where WR ≡ RAwj

1 + wA and WNR = τRBwj1 + wB.

Proposition 1 shows that under general conditions on the relative wages w andthe relative rates of return R, a greater number of highly skilled illegal migrantsare more likely to return.

Proposition 1 If the “normal” (percentage) wage gap w ≡ wB

wA is strictly higher

than the “normal” (percentage) rate-of-return gap R ≡ RB

RA , i.e.

w

R≡

wB

wA

RB

RA

> 1

10

then, net utility from return migration — therefore the probability of returning inthe home country — is an increasing function of the individual level of skills.

Proof.See Appendix B

This result is particularly important since it highlights how the effect of ille-gality as a brain waste, in both the labor market and the accession of financialmarkets, induces highly skilled migrants to leave the host country. While ourcurrent simple framework does not allow us to make general inferences regardingoverall welfare, it seems reasonable to assume that illegality costs the host country,as it induces the more productive individuals to leave first.

The net utility is also a decreasing function of the probability of legalization,as the first derivative of U j,∗ with respect to γ proves:

∂U j,∗

∂γ= −(1 + δ)(1− τ)wB

τRBwj1 + wB

As intuitively expected, better prospects for period 2 increase the expectedincome from staying in the host country and reduce the incentives to return.

These latter two results are the main objectives of the empirical analysis, pre-ceded by a presentation of the data set, in the following sections.

3 Individual Data on Irregular Immigrants: The

Survey on Illegal Migration in Italy

We use a unique source of data: the Survey on Illegal Migration in Italy (SIMI,henceforth). SIMI was collected from January to September 2003 by a team ofresearchers at the Department of Economics of the University of Bari with thesupport of AGIMI-Otranto. The outcome of this joint effort is a survey on themain demographic and socio-economic characteristics of a representative sampleof 920 illegal immigrants, as well as their motivations and future expectations.

By means of “illegal immigrant” (i.e. the sampling unit) we define a (at least18-year old) clandestine or asylum seeker that has been in Italy for a period nolonger than 6 months. This short period minimizes the measurement error wheninterviewees were asked to recall previous events. One of the aims of the survey isto obtain an accurate recollection of earnings and expenditures before migration, aswell as future expectations before departure. These immigrants were interviewed inthree types of centers, i.e. Center of Temporary Permanence, reception centers and

11

help centers spread in the four main regions mostly affected by the phenomenonof illegal entrance (Apulia, Sicily, Calabria and Friuli Venezia Giulia).

More precisely, the observational unit is identified according to the legal statusof the immigrants and in our study we consider the following four categories:

1. individuals applying for asylum or refugee status, i.e.:

• individuals under temporary protection for humanitarian aid;

• individuals that should be repatriated to a country where they wouldbe persecuted for reasons concerning race, gender, language, religion,opinions, citizenship, personal or social condition or that would be repa-triated to a country where they would not be protected from prosecution(ex art.19, 1◦ comma, D.lgs. no.286/98);

2. individuals waiting for a rejection decree with accompaniment to the closestborder ; the rejection decree is usually issued by the local police authority(Questore) to an individual that arrived in Italy avoiding border controlsand that was stopped immediately after her/his arrival;

3. individuals waiting for an expulsion decree: the decree is issued by the localadministrative authority (Prefetto) when the migrant avoided border controlsand was not yet rejected;

4. clandestine migrants, i.e. a foreigner with an expired (or no) visa that hasbeen in Italian territory for no longer than 6 months and that is present intypical migrant meeting points, like “soup kitchens”, orientation activitiesprovided by voluntaries and NGOs, etc.

The 920 individuals interviewed belonged to 55 different nationalities, withthe six largest fractions coming from Iraq (9.6%), Liberia (9%), Sudan (5.4%),Morocco (5.1%), Senegal (4.8%), Turkey (4.8%). The total number of individualsinterviewed represented 10.82% of all the 8,502 illegal migrants that were hostedin the selected centers in the period January-September 2003.

According to our data, the average illegal migrant entering Italy is young (about27 years old) and healthy. Most of the interviewees stated to be literate (85.8%),with some of them claiming a considerable level of schooling, although only about1/3 of them declared having a driving licence (35.2%). Nevertheless, about 70%of the interviewees indicated possessing low-skill qualifications.

Several socio-economic indicators were also measured by considering the “ge-ographical origin” within the country (whether coming from large cities or fromthe periphery or from the countryside), the availability of different utilities in the

12

original home, the occurrence of recent natural disasters and economic crisis in thearea of the migrant’s dwelling.

The declared individual monthly income in the country of origin was on averagearound 145 USD, with a very high variability due to the extreme heterogeneityof the socio-economic conditions of the interviewees. It is noteworthy that morethan half of the interviewees, once settled down in the country of final destination,expected to earn a monthly wage between 500 and 1,000 USD, with an average of937 USD.

Migration is a major investment for the family: on average it is equivalent to2 years of family earnings in the country of origin. Finally, it is worth remarkingthat 1/3 of the respondents judged their monthly income as “very volatile”.8

About 60% of the interviewees declared to have intentions to return homeand our empirical analysis is based on their characteristics in order to determinewhich variables affect more the decision to return home and whether they are inaccordance with the theoretical model presented in Section 2.

4 Empirical Analysis

In order to test the main implications of our theoretical model, we implement aprobit model on the intentions to return (i.e. the dependent variable is equal to1 if the individual expects to return home, zero otherwise). Definitions and basicstatistics of the explanatory variables are presented in Appendix C.

Following the theoretical model presented in Section 2, the main objectiveof this empirical analysis is to test whether the choice of returning home of anindividual that has illegally migrated is influenced by his or her skills (parameteraj) and by the possibility of being granted a legal status (parameter γ).

Here, we measure individual skills by means of three different variables: yearsof schooling, individual skills and qualifications and proficiency of the language ofthe intended country of destination. In accordance with the model of Section 2,for the more skilled migrants the brain waste effect associated with the status ofillegal migrant increases the opportunity cost of continuing to reside illegally in thedestination country. Thus, we expect variables related to skills to have a positiveeffect on the probability of returning in the country of origin.

Moreover, we proxy for the different probability of being granted legal statususing a dummy for clandestine. Generally speaking, illegal immigrants may bedivided into two broad categories: asylum seekers and clandestine immigrants.Asylum seekers are motivated to notify their presence to the authorities of thereceiving country, whereas clandestine immigrants shy away from official contacts

8For a detailed description of the sampling design, of the adopted questionnaire and of otherresults see Chiuri, et al. (2004).

13

and tend to live working quietly, waiting for the next amnesty which will makethem legal migrants. The probability of being granted legal status, while positivefor both categories of migrants, is generally higher for asylum seekers. Since beinglegal increases the ability of migrants to fully employ in the country of destinationher human and financial capital, we expect the effect of clandestine to be positiveon the propensity to return.

The willingness to return home also depends on expected economic opportuni-ties in the country of origin (i.e. the “normal” wage wA in model, or negatively thewage gap w). Return migration will be generally higher in countries that are at anintermediate level of development and would offer opportunities to migrants whohave accumulated human and financial capital. For this reason, we introduce twovariables for infrastructure in the country of origin, which are both expected tohave a positive effect on return: infrastructure (macro) that measures the relativeendowment of infrastructure at the country level and infrastructure (micro) that isa proxy for individual access to basic infrastructures at the level of the village/cityof origin. Expectations on future opportunities in the country of origin are alsoinfluenced by previous job experiences in the country of origin. Thus, we includea dummy variable for unemployed in the home country before migrating, which isexpected to have a negative influence on the probability of returning.

Moreover, illegal migrants might also find better employment opportunitiesvia migrants already established in the country of destination. In fact, thoseindividuals might provide the newcomers with information about labor marketopportunities and increase their probability of acceding to better-paid and stablejobs. In order to capture this effect, we include a dummy for migration network(migronetwork). In terms of our theoretical model, the existence of a networkimproves the ability of illegals to find a job. Therefore, we expect migronetworkto have a negative effect on the probability of return.

In order to have a complete empirical specification, we also introduce few vari-ables that control for the other factors that might affect the choice of returning.In fact, together with business and entrepreneurial motivations, one might decideto return to the country of origin because of family and cultural ties [see Dust-mann (2003a)].9 We therefore include a number of proxies that give a measure ofthe intensity of family ties, such as number of children, children in the destina-tion countries and relatives left at home. These are expected to have a positive,negative and positive effect on the return choice, respectively.10

9More broadly these factors might also proxy for the psychic cost of migration and may enterour model as a fixed disutility flow for each period the migrant is far away from the family. Anextension of the model of Section 2 is straightforward and is available from the authors uponrequest.

10See Dustmann (2003a) who highlights the importance of children in shaping parents’ returndecision.

14

As far as cultural ties with the country of origin are concerned, it is widelyaccepted that the costs of residing in a foreign country increase with the degree ofcultural and social diversity between the origin and destination countries. A dif-ferent religion is one important dimension on which such diversities are expressed.Hence, we include a dummy variable, Muslim, that aims to capture the, generallygreater, psychological cost of migration faced by individuals of Islamic religion,and this is supposed to have a positive effect on the return choice. Along the sameline, we include the (log of) geographical distance as a proxy for the monetaryand psychological cost of migration (when distance is short migrants can affordfrequent journeys back home) and previous migration experience, given that previ-ous moves generally lower the non-monetary and psychological costs of subsequentmigrations. These variables are expected to have a positive and negative effect onthe return choice, respectively.

Aside from our theoretical model, the peculiarity of our dataset also allowsus to analyze the effect of social conflict and financial or economic crisis in thevillage/city of origin on the choice of return.11 These events might have profoundand different implications on the intentions to return. In fact, while social conflictsor civil wars might have a permanent effect on migration, economic or financialcrisis might lead to a temporary out-migration which might be subsequently re-absorbed when economic conditions improve again.

Along a similar line of thinking, we control for the effects on return intentions ofbelonging to a minority religious and ethnic group in the home country. Minoritygroups in many countries of origin, which are represented in our sample, sufferfrom discrimination and sometimes violent persecutions. Hence, in our analysiswe include an interaction effect between a dummy variable minority and an indexof ethnic polarization, which aims to capture potential conflicts and concentrationof power “outside” the minority. This index ranges from 0 to 1 and polarizationreaches a maximum when there are two religious/ethnic groups of equal size.12 Asa matter of construction, this variable is expected to have a negative effect on theprobability of return.

Finally, we include a macro area dummy in order to capture the characteristicsof the geographical areas of origin that are not observable.13

Table 1 and 2 show the estimates and the relative marginal effects, respectively.Although we present the results of different specifications in what follows we only

11In terms of our model, they may be related once more to the “normal” wage in the countryof origin wA or negatively to the wage gap w, although the two variables will prove to have adifferent effect among each other.

12For recent analysis concerned with the effects of religious and ethnic polarization on economicdevelopment see Montalvo and Reynal-Querol (2003, 2004).

13The limited number of observations together with the large number of countries in our datasetdoes not allow us to use country dummies.

15

comment on the most completed one (Model 4).14

[Table 1 about here.]

[Table 2 about here.]

Results are generally in line with our expectations. Skills and education signif-icantly increase the probability of return to the home country. The probability ofreturn of a relatively skilled person is 9.7% higher than the probability of return ofan individual with no or low skills. Individuals with the lowest level of educationin the sample are 15% less likely to return than individuals with a higher level.Also, the knowledge of the language of the intended destination countries increasesthe likelihood to return by 11%.

These findings contrast with most existing studies on return migration andreturn intentions of legal migrants. Dustmann (1996, 2003b) using data fromthe German Socio-Economic Panel finds a negative effect of year of schooling onthe intention to return in the home country. However, the author finds that forthose who intend to return, schooling has a negative impact on the duration ofthe migration spell. In relation to this last result, Dustmann’s explanation is thathigher schooling, by guaranteeing higher salary, reduces the time needed to achievea pre-determined saving target. In a related study on the factors which affect thereturn migration of a cohort of foreign-born in the US, Reagan and Olsen (2000)find no evidence of skill bias in return migration. On the other hand, our resultsare in line with Zhao (2002)’s. In his analysis on rural to urban migration in China,the author finds that better educated and skilled rural migrants are more likelyto return to their village of origin. The explanation offered by the author fits ourinterpretation: both the strong labor segmentation in the urban labor market andthe tight migration regulatory system in China prevent the full participation ofskilled workers from rural area. This imposes heavy costs on skilled migrants interms of rewards to education and work experience.

As expected, the coefficient on the dummy for clandestine is positive and highlysignificant. The coefficient on migration network is significant and positive; thismight be due to the fact that the existence of established networks relativelyreduces the risks associated with the migratory experience.

Illegal migrants are also found to be more willing to return in countries thatare relatively more developed.15 Countries that have an above average level ofinfrastructures (as measured by the dummy infrastructure (macro) are 15% morelikely to attract migrants back home. As well, migrants who have declared to have

14Other specifications are also available from the authors upon request.15Let us recall that Proposition 2 in the Appendix underlines that all illegal migrants will go

back home when the wage gap is lower than the rate-of-return gap.

16

access to electricity and/or drinkable water in their home are 9.5% more likely toreturn [infrastructure (micro)].16 As was to be expected, we find that individualswho were unemployed back home are significantly less likely to return.

Most control variables have the expected sign. For instance, we find evidenceof the importance of family and cultural ties. In our estimations, an individualwith two children left in the home country is 16.8 percentage points more likelyto return than in the case where the children were already in the country of desti-nation. Also the size of the family left in the country of origin significantly affectreturn intentions. Our evidence is in conformity with Dustmann (2003a) wherethe presence of children in the host country negatively affects the return intentionof parents.

As expected, past migration experience reduces the probability of return plans.Also, the coefficients on the proxies for monetary and psychic cost of migration,namely distance and muslim, are significant and positive, respectively. Finally, wefind that illegal migrants from European and North African countries are morelikely to return than those coming from other countries.

Interestingly we also acknowledge that social conflicts and economic crisis haveopposite effects on the return choice. The effect of having experienced an economicor financial crisis in the village of origin seems to be temporary whereas socialconflicts have a more permanent effect on migration.17

5 Conclusions and Future Work

In this paper we assumed that the status of illegal migrant hinders the full utiliza-tion of individual skills. As a consequence, the opportunity cost of returning homeis lower for highly skilled migrants rather than individuals with few or no skills.

This result has been proved both theoretically and empirically. A simple two-period model with brain waste effect and probability of being granted legal statushas shown that the return choice is more likely for individuals with more abilities.A higher probability of being granted legal status (as it is for asylum seekers ratherthan clandestine immigrants) decreases the probability of returning home.

Empirical estimates of a probit model on the intentions to return home havebeen obtained on a sample of 920 illegal immigrants hosted in Italian centers. Both

16This variable might also be interpret as the level of relative deprivation of individuals in thehome country.

17We note this finding confirms the importance of a coordinated, timely and efficient inter-national conflict prevention activity. Also migrants belonging to a religious or ethnic minorityin the country of origin are less likely to return: the probability of remaining in the destinationcountry is increasing in the degree of religious polarization (i.e. the higher is potential hostilityfaced by a religious minority in the country of origin). See Chiuri et al. (2003).

17

skill measures (years of schooling, host-country language proficiency, level of skillof job at home) and proxies for higher probability of getting a legal status (likebeing an asylum seeker rather than a clandestine) affect the intentions to returnhome in the predicted direction. Other control variables prove the validity of theempirical model.

Since migration flows have proved to be unavoidable, the main message ofthis paper pinpoints the need to carefully design new immigration policies. Inparticular, it ought to be considered that a generic ban is not neutral and givesgreater incentives to the more skilled workers to return home rather than to thelow-skill migrants. An analysis of welfare considerations for both the host andthe home country would of course require a much richer theoretical model, thatwould include the effects of new entrants on the host labor markets, a multi-period framework and the possible interactions among natives, legal and illegalimmigrants. Another important extension of the model would regard the lengthof stay in the host country and whether this may also depend on the individual’sskills and the degree of illegality. All these extensions may be the task of futurework.

References

[1] Borjas G.J. and Bratsberg B. (1996), “Who leaves? The emigration of theforeign-born”, Review of Economics and Statistics, Vol 78 (1), pp. 165-67.

[2] Borjas, G., Bronars S. G., Trejo S. J. (1992), “Self-Selection and internalmigration in the United States”, Journal of Urban Economics, 32: 159-185.

[3] Chiswick Barry R. (2005), The Economics of Immigration, Edward ElgarPublishing Limited, Cheltenham, UK.

[4] Chiuri, M.C., G. De Arcangelis and G. Ferri (2003), “Crises in the Countriesof Origin and Illegal Immigration into Italy”, mimeo University of Bari.

[5] Chiuri, M.C., G. De Arcangelis, A. D’Uggento and G. Ferri (2004), “IllegalImmigration into Italy: Results from a Recent Field Survey”, CSEF WorkingPaper n. 121, (http://www.dise.unisa.it/WP/wp121.pdf).

[6] Dustmann C. (1996), “Return migration: the European experience”, Eco-nomic Policy, Vol. 22, pp. 214-250.

[7] Dustmann C. (1997), “Return Migration, uncertainty and precautionary sav-ings”, Journal of Development Economics, Vol. 52, pp. 295-316

18

[8] Dustmann C. (2003a), “Children and Return Migration”, Journal of Popula-tion Economics Vol.16, pp. 815-830.

[9] Dustmann C. (2003b), “Return migration, wage differentials, and the optimalmigration duration”, European Economic Review, Vol. 47, pp. 353-369.

[10] Galor O., Stark O. (1990), “Migrants savings, the probability of return mi-gration and migrants performance”, International Economic Review, Vol. 31,pp. 463-467;

[11] Jandl M. (2004), “The Estimation of Illegal Migration in Europe”, Studi Em-igrazione – Migration Studies, Vol. 51, pp 141-155.

[12] Montalvo J. G., Reynal-Querol M. (2003), “Religious polarization and eco-nomic development”, Economics Letters, Vol. 80, pp. 201-210.

[13] Montalvo J.G., Reynal-Querol M. (2004), “Ethnic Diversity and EconomicDevelopment”, Journal of development Economics, Vol. 76, pp. 293-323.

[14] Mesnard A. (2004), “Temporary migration and capital market imperfections”,Oxford Economic Papers, Vol. 56, pp. 242-262.

[15] Reagan P.B. and Olsen R. J. (2000), “You Can Go Home Again: Evidencefrom Longitudinal Data”, Demography, Vol. 37 (3), pp.339-350.

[16] Stark O. (1992), The Migration of Labor, Blackwell, Oxford.

[17] Zhao Y. (2002), “Causes and Consequences of Return Migration: RecentEvidence from China”, Journal of Comparative Economics, Vol. 30, pp. 376-394.

19

APPENDIX

A Sufficient Conditions on the Sign of the Net

Utility

Proposition 2 (Sufficient conditions for all migrants to stay in the host country B)No migrant decides to return, i.e. U j,∗ < 0, if:

(i) the “normal” (percentage) wage gap w ≡ wB

wA is strictly higher than the “nor-

mal” (percentage) rate-of-return gap R ≡ RB

RA :

w

R≡

wB

wA

RB

RA

> 1

(ii) both the “normal” wage and the “normal” rate of return are strictly higherin the host country rather than in the home country, i.e.

τwB > wA τRB > RA

Proof. Let us rewrite the net utility in eq. (8) as follows:

U j,∗ ≡ ln

[RAw1 + wA

τRBw1 + wB

τRB

RA

]+ δ ln

[RAw1 + wA

τRBw1 + wB

]or

U j,∗ ≡ ln

[τRBRAw1 + τRBwA

τRBRAw1 + RAwB

]︸ ︷︷ ︸

[1]

+ δ ln

[RAw1 + wA

τRBw1 + wB

]︸ ︷︷ ︸

[2]

(10)

The first term [1] is negative if (and only if):

τRBRAw1 + τRBwA < τRBRAw1 + RAwB

or

RAhwB

τRBwA> 1

which can be written in terms of wage gap and rate-of-return gap:

20

w

R≡

wB

wA

RB

RA

>τ

h

Let us recall that h ≡ [(1 − γ)τ + γ]; hence, the fraction τh

is certainly lowerthan 1 and the term [1] is always negative if

w

R> 1 (11)

The second term [2] is negative when:

RAw1 + wA < τRBw1 + wB

or

RAτajwB + wA < τRBτajwB + hwB

Let us rewrite the previous condition by employing the wage gap w and therate-of-return gap R:

RAwτaj (τR− 1)︸ ︷︷ ︸[A]

> (1− hw)︸ ︷︷ ︸[B]

(12)

A sufficient condition for (12) is that the term [A] is positive and the term [B]is negative.

This occurs when:

τR > 0 ⇒ R >1

τ⇒ τRB > RA

for term [A]; and:

hw > 1 ⇒ w >1

h

for term [B].Since 1

h< 1

τ, but condition (11) must be satisfied for term [1] to be negative,

then a sufficient condition on w would be:

w >1

τ⇒ τwB > wA

21

Proposition 3 (Sufficient conditions for all migrants to return the home country A)All migrants decide to return, i.e. U j,∗ > 0, if:

(i) the “normal” (percentage) wage gap w ≡ wB

wA is strictly lower than the “nor-

mal” (percentage) rate-of-return gap R ≡ RB

RA :

w

R≡

wB

wA

RB

RA

<τ

h

(ii) both the “normal” wage and the “normal” rate of return are strictly lower inthe host country rather than in the home country, i.e.

hwB < wA τRB < RA(< hRB)

Proof. Let us recall Eq. (10) from the proof of Proposition 2:

U j,∗ ≡ ln

[τRBRAw1 + τRBwA

τRBRAw1 + RAwB

]︸ ︷︷ ︸

[1]

+ δ ln

[RAw1 + wA

τRBw1 + wB

]︸ ︷︷ ︸

[2]

By following the same steps as for Proposition 2, the term [1] is now positiveif (and only if):

w

R≡

wB

wA

RB

RA

<τ

h

Hence, a necessary condition for the previous inequality to hold is:

w < R

since τ < h.

By using inequality (12) and the same steps as in Proposition 2, it is easy toshow that the term [2] is certainly negative if (and only if):

RAwτaj (τR− 1)︸ ︷︷ ︸[A]

< (1− hw)︸ ︷︷ ︸[B]

A sufficient condition such that the previous inequality holds is that term [A]is negative and term [B] is positive, which occurs respectively if:

τR < 1 ⇒ R <1

τ

22

and

hw < 1 ⇒ w <1

h

Since w must be lower than R for term [1] to be positive, then both conditionsare satisfied if:

1

h< R <

1

τand w <

1

h

B Proof of Proposition 1

Proof.When taking the first derivative of the net utility from return migration, we

obtain:

∂U j,∗

∂aj=

(1 + δ)τwB

WRWNR

(hRAwB − τRBwA

)where WR ≡ RAwj

1 + wA and WNR = τRBwj1 + wB.

The net utility is then strictly increasing in the skill level aj if and only if:

hRAwB > τRBwA

or:

w

R≡

wB

wA

RB

RA

>τ

[(1− γ)τ + γ]

Notice that, since γ is a probability, then h is a linear combination betweenτ (which is lower than 1) and 1. Hence, the fraction on the right-hand-side iscertainly lower than 1.

As a consequence, the condition:

w

R≡

wB

wA

RB

RA

> 1

is sufficient to assure that U j,∗ is increasing in aj.

23

CV

ari

able

sD

esc

ripti

on

Desc

ripti

on

Exp.

sign

Mean

St.

Dev

Type

Sourc

e

Yea

rsof

scho

olin

gnum

ber

ofye

arsth

ein

div

id-

ual

atte

nded

school

(cat

e-go

rica

lva

riab

le:

0non

e,1

from

1to

5ye

ars,

2fr

om6

to8

year

s,3

from

9to

11ye

ars,

4m

ore

than

12ye

ars)

+3.

371.

21ca

tego

rica

lQ

ues

tion

nai

re

Hos

t-co

untr

yla

ngu

age

profi

cien

cy

equal

s1

ifth

em

igra

nt

ispro

fici

ent

inth

ela

ngu

age

ofth

ein

tended

des

tinat

ion

countr

y(b

asic

orad

vance

dle

vel)

,0

other

wis

e

+0.

460.

49dum

my

Ques

tion

nai

re

Hig

hly

skilled

wor

ker

equal

s1

ifth

ein

div

id-

ual

pos

sess

eshig

hly

qual

-ifi

edjo

bsk

ills

(tra

nsl

ator

,se

cret

ary,

finan

cial

advis

or,

doct

oror

chem

ist,

law

yer,

teac

her

,m

anag

er,

consu

l-ta

nt,

entr

epre

neu

r)an

d0

other

wis

e

+0.

190.

39dum

my

Ques

tion

nai

re

Cla

nde

stin

eeq

ual

s1

ifth

ein

div

idual

de-

clar

esto

be

acl

andes

tine,

0ot

her

wis

e

-0.

30.

46dum

my

Ques

tion

nai

re

continued

onnex

tpa

ge

24

continued

from

prev

ious

page

Desc

ripti

on

Exp.

sign

Mean

St.

Dev

Type

Sourc

e

Mig

ro-

net

wor

keq

ual

s1

ifth

ein

div

idual

has

rela

tive

san

dfr

iends

inth

ein

tended

countr

yof

des

tina-

tion

,0

other

wis

e

+/-

0.18

0.39

dum

my

Ques

tion

nai

re

Infra-

stru

cture

(mic

ro)

equal

s1

ifth

ein

div

idual

has

elec

tric

ity

and/o

rdrinka

ble

wat

erin

the

hou

seof

resi

-den

cein

the

hom

eco

untr

y,0

other

wis

e

+0.

760.

43dum

my

Ques

tion

nai

re

Infra-

stru

cture

(mac

ro)

equal

s1

ifth

ein

div

id-

ual

com

esfo

rma

coun-

try

wher

eth

enum

ber

ofte

lephon

em

ainlines

,dai

lynew

spap

ers,

radio

and

tele

-vis

ions

sets

for

1.00

0peo

ple

ishig

her

than

the

aver

age

ofth

e56

countr

ies

inou

rsa

m-

ple

,0

other

wis

e

+0.

160.

37dum

my

Cou

ntr

yTa-

ble

sIT

Cat

glan

ceD

evel

-op

men

tD

ata

Gro

up,W

orld

Ban

k

Unem

ploy

edin

the

hom

eco

untr

y

equal

s1

ifth

em

igra

ntis

not

emplo

yed

bef

ore

dep

artu

re,

0ot

her

wis

e

-0.

560.

5dum

my

Ques

tion

nai

re

Soc

ialco

n-

flic

teq

ual

s1

ifth

em

igra

nt

de-

clar

esth

atth

ere

was

aso

-ci

alco

nflic

tin

the

villa

geor

city

ofor

igin

(res

iden

ce)

inth

ela

st5

year

s,0

other

wis

e

-0.

680.

46dum

my

Ques

tion

nai

re

continued

onnex

tpa

ge

25

continued

from

prev

ious

page

Desc

ripti

on

Exp.

sign

Mean

St.

Dev

Type

Sourc

e

Eco

nom

iccr

isis

equal

s1

ifth

em

igra

nt

de-

clar

esth

atth

ere

was

anec

o-nom

icor

finan

cial

cris

isin

the

villa

geor

city

ofor

i-gi

n(r

esid

ence

)in

the

last

5ye

ars,

0ot

her

wis

e

+/-

0.86

0.34

dum

my

Ques

tion

nai

re

Min

ority

×et

hnic

pola

r-izat

ion

inde

x

inte

ract

ion

effec

tbet

wee

nth

edum

my

vari

able

“mi-

nor

ity”

(equal

to1

inth

ein

-div

idual

sbel

ongs

toa

reli-

giou

sm

inor

ity

inth

eco

un-

try

ofor

igin

)an

dth

eIn

dex

ofE

thnic

Pol

ariz

atio

n(s

eeM

onta

lvo

J.G

.an

dR

eynal

-Q

uer

olM

.,20

04)

whic

hca

p-

ture

sth

edeg

ree

ofet

hnic

pol

ariz

atio

nw

ithin

aco

un-

try

(index→

0m

eans

very

hig

hpol

ariz

atio

n;

index

→1

mea

ns

very

low

pol

ariz

a-ti

on)

-0.

230.

34dum

my

CIA

-T

he

Wor

ldFac

t-book

and

Ques

tion

-nai

re

N.

ofch

il-

dren

num

ber

ofch

ildre

n-

0.6

1.1

continuou

sQ

ues

tion

nai

re

Chi

ldre

nin

the

dest

inat

ion

countr

y

equal

s1

ifon

eor

mor

ech

il-

dre

nar

ein

the

inte

nded

des

-ti

nat

ion

countr

y(n

otnec

es-

sari

lytr

avel

ing

with

the

re-

spon

den

t),0

other

wis

e

-0.

060.

24dum

my

Ques

tion

nai

re

continued

onnex

tpa

ge

26

continued

from

prev

ious

page

Desc

ripti

on

Exp.

sign

Mean

St.

Dev

Type

Sourc

e

Relat

ives

inth

eho

me

countr

y

num

ber

ofre

lati

ves

that

are

left

inth

eco

untr

yof

orig

in-

5.17

3.98

continuou

sQ

ues

tion

nai

re

Pas

tm

i-gr

atio

neq

ual

s1

ifth

ein

div

idual

has

alre

ady

don

ea

mig

ra-

tion

exper

ience

(inte

rnal

orin

tern

atio

nal

),0

other

wis

e

-0.

270.

44dum

my

Ques

tion

nai

re

Dista

nce

(in

log)

(log

ofth

e)dis

tance

inK

m(G

eogr

aphic

alco

-ord

inat

esw

her

euse

dto

calc

ula

tedis

-ta

nce

;ro

unded

lati

tude

and

longi

tude

figu

res

wer

euse

dfo

rth

epurp

ose

offindin

gth

eap

pro

xim

ate

geog

raphic

cente

rof

the

orig

inan

ddes

-ti

nat

ion

countr

ies)

+/-

7.95

0.7

continuou

sG

azet

teer

ofC

onve

nti

onal

Nam

es,T

hir

dE

dit

ion,

Au-

gust

1988

,U

SB

oard

onG

eogr

aphic

Nam

esan

don

other

sourc

es.

Muslim

equal

s1

ifth

ein

div

idual

de-

clar

esto

be

aM

usl

im;0

oth-

erw

ise

-0.

590.

49dum

my

Ques

tion

nai

re

Euro

peeq

ual

s1

ifth

ein

div

idual

countr

yof

orig

inis

inE

u-

rope,

0ot

her

wis

e

+0.

250.

43dum

my

Ques

tion

nai

re

Nor

thA

frica

equal

s1

ifth

ein

div

idual

countr

yof

orig

inis

inN

orth

Afr

ica

(Afr

o-M

editer

ranea

nco

untr

ies)

,0

other

wis

e

+0.

110.

31dum

my

Ques

tion

nai

re

continued

onnex

tpa

ge

27

continued

from

prev

ious

page

Desc

ripti

on

Exp.

sign

Mean

St.

Dev

Type

Sourc

e

Asia

equal

s1

ifth

ein

div

idual

countr

yof

orig

inis

inA

sia,

0ot

her

wis

e

-0.

270.

44dum

my

Ques

tion

nai

re

Afr

ica

(exc

lN

orth

Afr

ica)

equal

s1

ifth

ein

div

id-

ual

countr

yof

orig

inis

inA

fric

a(e

xcl

uded

Nor

thA

fric

anco

untr

ies)

,0

other

-w

ise

-0.

370.

48dum

my

Ques

tion

nai

re

Am

erica

equal

s1

ifth

ein

div

idual

countr

yof

orig

inis

inC

en-

tral

and

Lat

inA

mer

ica,

0ot

her

wis

e

-0.

005

0.07

dum

my

Ques

tion

nai

re

28

Table 1: Estimates of the Probit Model: Different Spec-ifications

Regressors Model 1 Model 2 Model 3 Model 4Years of sch. 0.110∗∗ 0.126∗∗ 0.119∗∗ 0.100∗∗

(0.036) (0.042) (0.045) (0.046)Host-c. lang. prof. 0.514∗∗ 0.399∗∗ 0.312∗∗ 0.294∗∗

(0.089) (0.104) (0.112) (0.112)Highly skilled worker 0.268∗

(0.141)Clandestine 0.636∗∗ 0.657∗∗

(0.139) (0.140)Migronetwork 0.435∗∗ 0.421∗∗

(0.159) (0.159)Infrastr. (micro) 0.216∗ 0.249∗ 0.248∗

(0.122) (0.128) (0.128)Infrastr. (micro) 0.287∗ 0.46∗∗ 0.438∗∗

(0.153) (0.164) (0.165)Unemployed at home −0.199∗∗ −0.219∗∗ −0.227∗∗

(0.102) (0.107) (0.108)Social conflict −0.399∗∗ −0.285∗∗ −0.292∗∗

(0.127) (0.140) (0.140)Economic crisis 0.566∗∗ 0.55∗∗ 0.588∗∗

(0.148) (0.158) (0.159)Minority*eth. pol. −0.376∗∗ −0.323∗∗ −0.298∗

(0.149) (0.158) (0.159)N. of children 0.136∗∗ 0.145∗∗ 0.143∗∗

(0.050) (0.051) (0.051)Children in host c. −0.498∗∗ −0.446∗ −0.45∗

(0.223) (0.242) (0.243)Relatives at home 0.043∗∗ 0.042∗∗ 0.042∗∗

(0.013) (0.014) (0.014)Past migration −0.307∗∗ −0.23∗∗ −0.251∗∗

(0.109) (0.114) (0.115)Distance(in log) 0.427∗∗ 0.535∗∗ 0.531∗∗

(0.128) (0.137) (0.137)Muslim 0.304∗∗ 0.36∗∗ 0.368∗∗

(0.107) (0.115) (0.115)Asia −0.830∗∗ −0.794∗∗ −0.807∗∗

(0.218) (0.231) (0.231)continued on next page

29

Table 1: continued

Regressors Model 1 Model 2 Model 3 Model 4Africa (excl North) −0.522∗∗ −0.406∗ −0.41∗

(0.217) (0.235) (0.235)America −0.628 −0.737 −0.684

(0.683) (0.781) (0.781)Constant −0.340∗∗ −3.852∗∗ −5.138∗∗ −5.101∗∗

(0.128) (0.955) (1.038) (1.040)

Observations 866 798 752 752Pseudo R2 0.043 0.144 0.190 0.1932Log likelihood −556.83 −457.72 −410.56 −408.73Standard errors in parentheses / Probability of return (baseline) = 0.633* significant at 10%; ** significant at 5%

30

Table 2: Marginal Effects of Model (4)

Regressors Marg. Eff. Prob. (1)Years of schooling 0.038∗∗ 0.046

(0.017)Host-country lang. proficiency 0.110∗∗

(0.041)Highly skilled worker 0.097∗∗

(0.049)Clandestine 0.231∗∗

(0.045)Migronetwork 0.149∗∗

(0.052)Infrastracture (micro) 0.095∗

(0.05)Infrastracture (macro) 0.154∗∗

(0.053)Unemployed in the home country −0.085∗∗

(0.04)Social conflict −0.107∗∗

(0.05)Economic crisis 0.230∗∗

(0.062)Minority*ethnic polarization index −0.112∗ -0.037

(0.06)N. of children 0.054∗∗ 0.06

(0.019)Children in the destination country −0.176∗∗

(0.096)N. of relatives in the home country 0.016∗∗ 0.063

(0.052)Past migration −0.096∗∗

(0.045)Distance(in log) 0.199∗∗ 0.141

(0.052)Muslim 0.14∗∗

(0.043)Asia −0.31∗∗

(0.086)Africa (excl North Africa) −0.156∗

continued on next page

31

Table 2: continued

Regressors Marg. Eff. Prob. (1)(0.089)

America −0.268(0.294)

Probability of return (baseline) = 0.633(1) change in predicted probability as X changes of onestandard deviation centered around the mean valueStandard errors in parentheses* significant at 10%; ** significant at 5%

32